Last time I discussed the “kinds of SETI” I laid out in my new paper on the arXiv. This time, I’ll discuss a nifty plot I made about upper limits.

At a workshop at the Keck Institute for Space Science a couple of years ago, I put together a graphic describing how I think we should think about placing upper limits in SETI:

The idea is that if you do a search and find nothing, you need to let people know what it is that you did not find so that we can chart progress, draw conclusions, and set markers down for what the next experiment should look like. My analogy is the dark matter particle detection community, which (similarly to SETI) must solve the funding paradox of using a lack of results to justify continued funding.

The idea is that you have some parameter that marks the strength of a purported (ambiguous) technosignature (like waste heat for Dyson Spheres).  If you perform, say, a search of all-sky data along that axis, then you will end up with many potential detections. Virtually all of these are (presumably) natural sources, so what can you do?

Well, the easiest first step is that one of the sources is the strongest, meaning that you instantly have an upper limit: no technosignatures exist above that threshold. If you’re the first to interpret the data that way, then you’ve made progress!  We now know something we didn’t before.

Then the sleuthing kicks in.  What are those top, say, 10 sources?  Are they all well-known giant stars with dusty atmospheres and certainly not Dyson Spheres?  If so, then you’ve just lowered the upper limit.

As you work your way down, your upper limit keeps improving, and you keep learning about what’s not out there. You also learn what you need to do to weed out all of the false positives to get to more meaningful and stringent upper limits.

This works for almost any technosignature with many confounders: structures on the moon, transiting megastructures, “salted” stars.

This formalism even works for machine learning-based anomaly detection, although in that case it might be hard to translate your upper limit into something physically meaningful, because the mapping between an anomaly score and the characteristic of the technology that would give that score might be obscure.

Next up: advice!